What are Diffusion Models?

<!-- Diffusion models are a new type of generative models that are flexible enough to learn any arbitrarily complex data distribution while tractable to analytically evaluate the distribution. It has been shown recently that diffusion models can generate high-quality images and the performance is competitive to SOTA GAN. --> <p><span class=&#34;update&#34;>[Updated on 2021-09-19: Highly recommend this blog post on <a href=&#34;https://yang-song.github.io/blog/2021/score/&#34;>score-based generative modeling</a> by Yang Song (author of several key papers in the references)].</span><br/> <span class=&#34;update&#34;>[Updated on 2022-08-27: Added <a href=&#34;#classifier-free-guidance&#34;>classifier-free guidance</a>, <a href=&#34;#glide&#34;>GLIDE</a>, <a href=&#34;#unclip&#34;>unCLIP</a> and <a href=&#34;#imagen&#34;>Imagen</a>.</span><br/> <span class=&#34;update&#34;>[Updated on 2022-08-31: Added <a href=&#34;#ldm&#34;>latent diffusion model</a>.</span><br/> <span class=&#34;update&#34;>[Updated on 2024-04-13: Added <a href=&#34;#prog-distll&#34;>progressive distillation</a>, <a href=&#34;#consistency&#34;>consistency models</a>, and the <a href=&#34;#model-architecture&#34;>Model Architecture section</a>.</span></p>

Open original